An experimental investigation into the syntax of
negation in Japanese*
Chung-hye Han, Dennis Ryan Storoshenko and Yasuko Sakurai(Simon Fraser University)
The goal of this paper is to provide a unified syntactic analysis of the two types of negation
in Japanese: regular negation with -na affixed onto the main verb, and wa-negation with the
so-called topic marker -wa on the main verb followed by auxiliary suru (‘do’) inflected with
-na. We establish that -na heads a projection of its own, and argue that this negation
projection is located low in the clause structure, within the lexical vP domain. Our proposal
is supported by the data concerning scope of negation and argument quantified NPs,
extracted through a set of psycholinguistic experiments using the Truth Value Judgment
Task.
Keywords: syntax of negation, scope, Truth Value Judgment Task, verb-raising, Japanese
1 Introduction
Japanese has two main ways of forming sentential negation: regular negation with -na
affixed onto the main verb, as in (1a), and wa-negation with the so-called topic marker
-wa on the main verb followed by auxiliary suru (‘do’) inflected with -na, as in (1b).1
(1) a. Ken-ga orenji-o shokudoo-de tabe-na-katta.
Ken-NOM orange-ACC cafeteria-in eat-NEG-PST
‘Ken did not eat an orange in the cafeteria.’
b. Ken-ga orenji-o shokudoo-de tabe-wa shi-na-katta.
Ken-NOM orange-ACC cafeteria-in eat-TOP do-NEG-PST
‘Ken did not eat an orange in the cafeteria.’
* We thank the three anonymous reviewers for their insightful comments and criticisms that were
crucial in improving this paper. Parts of this paper were presented at the 30th Annual Meeting of the
Berkeley Linguistics Society in 2004, the 1st International Conference on East Asian Linguistics in
Toronto, 2006, and the 2007 International Conference on Linguistics in Korea. We thank the audiences for
helpful comments and questions. All remaining errors are ours. This work was supported by SSHRC 410-
2003-0544 and SSHRC 410-2007-2169 to Han.
1 We gloss -wa as TOP (‘topic marker’) for lack of a better term, but stay neutral as to the discourse
function of this marker.
The goal of this paper is to provide a unified syntactic analysis of the two types of
negation. To this end, we will first establish that -na heads a projection of its own, and
then establish where this negation projection is located in the clause structure, using data
concerning scope of negation and argument quantified NPs. The data is extracted through
psycholinguistic experimentation using a technique known as the Truth Value Judgment
Task (TVJT), as in Crain and Thornton (1998). We will argue that the findings from our
experiments support a view that the negation projection is placed low in the clause
structure, within the lexical vP domain.
This paper is organized as follows. In section 2, we consider lexical and syntactic
approaches to Japanese negation and conclude that the data calls for a syntactic approach
with a negation projection (NegP) in the clause structure. We then argue that the scope
interaction between negation and argument quantified NPs (QPs) makes a good test for
the position of NegP in Japanese. We however point out that the data that is presented in
the extant literature on this topic cannot be used as it is because the scope judgments
reported there conflict with each other, making it impossible to reach any coherent
conclusions. This raises a question as to the validity of the methodology employed in
extracting these scope judgments. In section 3, we present a TVJT experiment that we
conducted in order to extract more reliable scope judgments. The test sentences in this
experiment contained negation and numeral QPs. After a discussion of the particulars of
our experiment, we present our findings followed by their implications for the syntax of
negation in Japanese. In section 4, we discuss a previously published experiment on the
scope of negation and universal QPs (Han et al., 2004) and show that the findings from
the universal quantifier experiment and the numeral quantifier experiment are consistent
with each other.
2 Negation projection
2.1 Motivating a negation projection
In principle, negation -na can be incorporated into the clause structure in two alternative
ways: lexically or structurally. Under the lexical approach, -na is a derivational
morpheme, entering the syntactic derivation as a part of a single syntactic unit along with
the verb to which it is affixed. Under the structural approach, -na is an inflectional
morpheme, projecting its own functional projection (NegP), and combines with the verb
through a morpho-syntactic operation, either verb-raising or INFL-lowering. The form of
the tense inflection on the verb occurring with -na appears to support the lexical
approach. In Japanese, the tense inflection on the adjective is distinct from the verb. For
example, while the past tense morpheme on verb tabe ‘eat’ is -ta as in (2a), the past tense
morpheme on the adjective oki ‘big’ is -katta as in (2b). It turns out that negated verbs
pattern with adjectives, not with verbs, taking the adjective tense inflection, as in (2c).
(2) a. Jiroo-ga pizza-o tabe-ta.
Jiroo- NOM pizza-ACC eat-PST
‘Jiroo ate pizza.’
b. Hon-wa oki-katta.
book-TOP big-PST
‘The book was big.’
c. Jiroo-ga piza-o tabe-na-katta.
Jiroo-NOM pizza-ACC eat-NEG-PST
‘Jiroo did not eat pizza.’
This patterning of tense inflection seems to suggest that -na is a derivational morpheme
that derives an adjectival category. However, there are at least two facts that support the
structural approach to -na.
First, syntactically, negated verbs still behave like verbs. For example, transitive
verbs usually assign an accusative case to their internal arguments as in (2a), while
transitive adjectives usually assign a nominative case to them as in (3). The internal
argument in negated verbs bear accusative case, patterning as a verb, not as an adjective,
as in (2c).
(3) Noriko-wa Shuya-ga suki desu.
Noriko-TOP Shuya-NOM fond of be
‘Noriko is fond of Shuya.’
Second, -na triggers weak island effects. While object extraction from an embedded
clause across a negative matrix predicate is possible as in (4a), adjunct extraction in the
same context is not possible (Hoji, 1985; Miyagawa, 2002). In (4b), only the reading in
which the time of Shuya’s thinking is questioned is available.
(4) a. Nani-oi Hanako-ga [Taroo-ga ti kat-ta to] omottei-na-katta-no?
what-ACCi Hanako-NOM Taroo-NOM ti buy-PST COMP think-NEG-PST-Q
‘What didn’t Hanako think that Taroo bought?’
b. Shuya-ga itsu Noriko-ga suupu-o non-da to omo-ana-katta-no?
Shuya-NOM when Noriko-NOM soup-ACC drink-PST COMP think-NEG-PST-Q
‘When didn’t Shuya think that Noriko drank soup?’
According to Rizzi (1990), an empty operator in [Spec,NegP] is responsible for weak is-
land effects in negative sentences. The fact that the same type of weak island effect is
found in Japanese negative sentences thus provides evidence for an empty operator in
[Spec,NegP], which in turn is evidence for positing a phrasal projection for negation.
2.2 Using scope to locate the negation projection
We have so far established that Japanese negation projects a NegP. The question now is
where in the clause structure NegP is projected, and whether there are two different posi-
tions for NegP, given that Japanese has two ways of forming sentential negation, regular
and wa-negation, as was illustrated in (1). The fact that these two negations cannot be
combined to form a double negation, as in (5), suggests that there is only one NegP
position in the Japanese clause structure.2
(5) *Ken-ga orenji-o shokudoo-de tabe-na-wa shi-na-katta.
Donald-NOM orange-ACC cafeteria-in eat-NEG-TOP do-NEG-PST
‘Ken didn’t not eat an orange in the cafeteria.’
The issue of determining the exact position of NegP turns out to be a difficult one,
even though there is a good diagnostic for it, namely scope interaction between negation
and an object QP. Two independently motivated background facts about Japanese syntax
make scope facts informative: object raising and scope rigidity.
First, the object NP in Japanese can occur before an adverbial or a postpositional
phrase, as in (1a), repeated below as (6). This means that in sentences such as (6), the
object is forced to undergo a clause-internal scrambling to a position higher than the
adverbial phrase. Assuming that the adverbial phrase is adjoined to vP, the object in (6)
can raise out of VP and adjoin to the higher layer of the vP, as in (7).
(6) Ken-ga orenji-o shokudoo-de tabe-na-katta.
Ken-NOM orange-ACC cafeteria-in eat-NEG-PST
‘Ken did not eat an orange in the cafeteria.’
2 A reviewer correctly points out that it is possible for Japanese sentences to contain multiple negations,
one contained within a -wa marked construction, and another sentence-finally, as in (i).
(i) Jiroo-ga piza-o tabe-na-i-no-de-wa na-i.
Jiroo-NOM pizza-ACC eat-NEG-NPST-NL-COP-TOP neg-NPST
‘Jiroo doesn’t not eat pizza.’
Note however that the second instance of negation in this sentence stands alone, bearing a separate tense
marker, which suggests a bi-clausal structure for constructions of this type. The status of this na is
discussed in Oishi (1986), and Storoshenko (2004) argues that it is not syntactically related to the regular or
-wa forms of negation.
(7)
Second, scope for argument NPs is fixed before, and not at, LF in Japanese, as inde-
pendently motivated by Scope Rigidity effects (Kuno, 1973; Kuroda, 1979; Hoji, 1985).
That is, the scope of argument QPs is determined by their surface c-command
relationship, without recourse to QR or reconstruction. Thus, a sentence with canonical
SOV word order containing quantifiers in both subject and object NPs only exhibits the
reading in which the subject scopes over the object, as in (8).
(8) Dareka on’nanoko-ga otokonoko daremo-o hihanshi-ta.
some girl-NOM boy every-ACC criticize-PST
‘Some girl criticized every boy.’ ( some>every, *every>some)
Putting the two background facts together, a prediction emerges. Starting with the
clause structure in (7), there are two possible positions for NegP in principle: NegP could
be low in the clause structure within vP as in Hypothesis 1 in (9a), or it could be higher in
the clause structure above vP as in Hypothesis 2 in (9b).
(9) a. Hypothesis 1 b. Hypothesis 2
If scope rigidity holds between negation and an argument QP, just as it holds between
argument QPs, then in a negative sentence with an argument QP, the relative scope of
negation and the QP will directly reflect their relative positions with respect to each
other.3
This means that in a negative sentence with a subject QP, under both Hypothesis 1
and Hypothesis 2 in (9), the subject QP should scope over negation but not the other way
around. Results from our experimental investigation confirm this expectation, as will be
discussed in section 3, supporting that negation has a scope freezing property. Given this,
we can use negative sentences with a vP-adjoined object QP to tease apart the two
hypotheses in (9), as they make different predictions: Hypothesis 1 predicts that the
object QP should scope over negation, and Hypothesis 2 predicts that negation should
3 A reviewer asks whether scopal relations between negation and NPIs will be informative. Following the
analysis of Nakanishi (2006), Japanese NPIs are in fact scalars which are interpreted as NPIs in negated
contexts, despite taking scope over negation at LF. Because Japanese NPIs do not need to fall under the
scope of negation to be licensed in this manner, NPI licensing will not be informative in determining the
scope of negation in Japanese.
scope over the object QP. The predictions are clear, but unfortunately, the claims made in
the literature about this topic are not.
2.3 Disagreement in the literature
When examining the relationship between sentential negation and an object QP, there are
ultimately three possible relations: neg>Q, Q>neg, or ambiguity in which, for a given
sentence, either relation can hold. In examining the existing literature on the topic, all
three of these positions can be found. Kuno (1980) enters this discussion with the claim
that verbally-adjoined negation takes scope only over the immediately preceding verb,
but amends this claim when examining a case with an object QP:
(10) Pai-o zenbu tabe-rare-na-katta.
Pie-ACC all eat-can-NEG-PST
‘I could not eat all the pie.’
For examples like (10), Kuno claims that the reading should primarily be neg>Q, despite
his earlier claim that negation should take the narrowest possible scope. The alternative
reading of Q>neg is given as a possible secondary reading. The notion that an object QP
should take scope over negation arises in Yatabe (1996), where it is claimed that all
quantifiers in Japanese should take scope over negation. In this analysis, readings of
neg>Q are never attained.
The possibility of outright ambiguity emerges in Ota and Kato (1986), where
examples such as (10) are given as equally neg>Q or Q>neg, without hedges such as
referring to primary or secondary readings. In attempting to solve a separate syntactic
puzzle, Miyagawa (2001) uses both subject and object QP scope judgments as a source of
evidence. Here, while the overall appraisal seems to agree with Kuno in that neg>Q
seems to be the preferred reading when encountering an object QP, Miyagawa notes that
for some consultants, Q>neg was also possible. Thus, while Kuno and others seem to
believe that negative sentences are subtly ambiguous, Miyagawa introduces the notion
that they may only be so for some native speakers.
There is much less said in the literature on the subject of the relative scope between
an object QP and -wa-negation. This being the case, what little there is to be found is
noteworthy in that, unlike the discussion of regular negation’s scope, all discussions of -
wa-negation’s scope are consistent. There is general agreement with the stance taken in
Ota and Kato (1986) that in -wa-negation, negation should always scope over the object
QP.
3 The numeral quantifier experiment
Given the conflicting claims in the literature on the scope judgments, we cannot use them
as they are to make any conclusions as to which hypothesis is superior. What could be the
source of the disagreement in the literature? One possibility might be a methodological
problem in the elicitation of scope judgments. Some speakers may have difficulty in
identifying a reading associated with a sentence without a sufficient discourse context.
Another possibility might be that the disagreement actually reflects speakers’ grammars.
It is possible that different speakers have different grammars, resulting in the apparent
disagreement in scope judgments. To address these issues, we designed a
psycholinguistic experiment using the TVJT technique and extracted scope judgments
that we think clearly reflect speakers’ grammars.
3.1 Methodology
The TVJT involves two experimenters. One experimenter acts out short stories in front of
the participant using toys. The other experimenter plays the role of a puppet (e.g., Mickey
Mouse) who watches the scenario alongside the participant. At the end of the story, the
puppet makes a statement about the story. The participant’s task is to determine whether
the puppet understood the story and say whether it told the truth or not.
This task provides rich discourse contexts in a simple method, with little memory
load on the participants. It has been shown to work in several languages such as English
and Kannada (Lidz and Musolino, 2002), Modern Greek (Papafragou and Musolino,
2003) and Korean (Han et al., 2007), and to work with both adults and children as young
as 4 years old (Crain and McKee, 1985; Crain and Thornton, 1998; Lidz and Musolino,
2002; Han et al., 2007). It is an ideal tool for obtaining subtle interpretive judgments,
which have proven to be difficult to obtain with a traditional method where native
speakers are presented with a set of sentences taken out of context.
For instance, to test how speakers interpret a sentence with negation and an object
QP, such as Cookie Monster didn’t eat every cookie, an experimenter enacts a story,
using Cookie Monster and three cookies. Cookie Monster is hungry and finds three
cookies. He eats only two of them. After the enactment is over, the puppet says, “I know
what happened. Cookie Monster didn’t eat every cookie. Am I right?” In this story, the
reading in which negation scopes over the object QP (neg> ) is true, and the reading in
which the object QP scopes over negation ( >neg) is false. So if a participant accepts the
test sentence in this context, then we can conclude that the grammar makes available the
wide scope reading of negation. But if a participant rejects the test sentence, we can
conclude that it must be because the grammar does not generate the neg> interpretation.
3.2 Participants
We tested 256 adult speakers of Japanese on the scope of negation and numeral QPs. The
participants were 20-30 year old native Japanese speakers living in Vancouver at the time
of the testing, who had spent no more than a combined span of 12 months in any English-
speaking country (including their residence in Canada at the time of the experiment).
3.3 Experimental design
The overall design of our experiment and the stories we used in the TVJT are closely
modeled after the work reported in Han et al. (2007). We designed the experiment to test
three factors with two levels each: grammatical function (Subject QP vs. Object QP),
negation (regular neg vs. -wa-neg) and scope (2>neg vs. neg>2). The experiment was
thus divided into 8 different conditions, each condition testing for the 2>neg or neg>2
reading in sentences containing regular neg or -wa-neg and a numeral quantifier futa
(‘two’) in the subject or the object. We used a between-subjects design, and randomly
assigned 32 participants to each condition. The design is summarized in Table 1.
Grammatical Function Scope Regular neg Wa-neg
Subject QP 2>neg n=32 n=32
neg>2 n=32 n=32
Object QP 2>neg n=32 n=32
neg>2 n=32 n=32
Table 1: Experimental design (Numeral quantifier study)
3.4 Materials
We used four different types of test sentences: a subject QP and regular negation as in
(11), a subject QP and -wa-negation as in (12), an object QP and regular negation as in
(13) and an object QP and -wa-negation as in (14). Each test sentence was used in a
scenario compatible with the 2>neg reading, and a scenario compatible with the neg>2
reading.
(11) Futa-ri-no otokonoko-ga suika-o shokudoo-de tabe-na-katta.
two-CL-GEN boy-NOM watermelon-ACC cafeteria-in eat-NEG-PST
‘Two boys did not eat a watermelon in the cafeteria.’
(12) Futa-ri-no otokonoko-ga suika-o shokudoo-de tabe-wa shi-na-katta.
two-CL-GEN boy-NOM watermelon-ACC cafeteria-in eat-TOP do-NEG-PST
‘Two boys did not eat a watermelon in the cafeteria. ‘
(13) Junko-ga futa-ri-no otokonoko-o rooka-de kera-na-katta.
Junko-NOM two-CL-GEN boy-ACC hallway-in kick-NEG-PST
‘Junko did not kick two boys in the hallway.’
(14) Junko-ga futa-ri-no otokonoko-o rooka-de kera-wa shi-na-katta.
Junko-NOM two-CL-GEN boy-ACC hallway-in kick-TOP do-NEG-PST
‘Junko did not kick two boys in the hallway.’
For instance, to test whether (11) and (12) can have the 2>neg reading, we used a
scenario with four boys, in which two boys each eat a watermelon, but the other two do
not. To test whether they can have the neg>2 reading, we used a scenario with two boys,
in which only one boy eats a watermelon. The pictures in Figure 1 are the scenes at the
end of the two scenarios. To test whether (13) and (14) can have the 2>neg reading, we
used a scenario with four boys and one girl named Junko. Junko is showing off her karate
moves. She kicks two boys, but does not kick the other two boys. To test the neg>2
reading with (13) and (14), we used a scenario with two boys, and one girl named Junko.
Junko kicks one boy, but does not kick the other. The pictures of the scenes at the end of
the two scenarios are given in Figure 2.
2>neg = True; neg>2 = False 2>neg = False; neg>2 = True
Figure 1: Scenarios for ‘Two boys did not eat a watermelon.’
2>neg = True; neg>2 = False 2>neg = False; neg>2 = True
Figure 2: Scenarios for ‘Junko did not kick two boys.’
Each participant was given four test trials. The test sentences stated by Mickey Mouse
in the eight conditions are given in Appendix A. In addition to the four test trials, each
participant was given four filler trials. The purpose of the filler trials was to separately
test the participants’ comprehension of negation and of numeral futa, and to prevent any
priming effects. To add some variety to the material, participants in regular negation
condition were given two filler sentences with -wa-negation as in (15a), and those in -wa-
negation condition were given two filler sentences with regular negation as in (15b).
(15)a. Kaba-ga kirikabu-ni nobori-wa shi-na-katta.
hippo-NOM tree trunk-onto climb-TOP do-NEG-PST
‘The hippo did not climb onto the tree trunk.’
b. Kaba-ga kirikabu-ni nobori-na-katta.
hippo-NOM tree trunk-onto climb-NEG-PST
‘The hippo did not climb onto the tree trunk.’
In addition, the participants in subject conditions were given two filler sentences with futa
in the object as in (16a), while those in object conditions were given two filler sentences
with futa in the subject as in (16b).
(16)a. Lisa-ga futa-ri-no otokonoko-o ie-no mae-de dakishime-ta.
Lisa-NOM two-CL-GEN boy-ACC house-GEN front-in hug-PST
‘Lisa hugged two boys in front of her house.’
b. Futa-ri-no otoko-ga neko-o omise-de nade-ta.
two-CL-GEN man-NOM cat-ACC shop-in pet-PST
‘Two men petted the cat in the shop.’
To introduce variation in the answers, we set up the filler scenarios such that two of the
fillers should be true and the other two should be false. The filler statements made by
Mickey Mouse in each condition are given in Appendix B.4
3.5 Procedure
The participants were tested in groups of 4 to 6 in a classroom. They were shown a series
of video clips of the scenarios described in section 3.4. They were first introduced to the
task with 4 practice trials. They were then given 4 test trials and 4 filler trials in a
pseudorandom order. They were each given a score sheet and were instructed to indicate,
for each trial, whether Mickey Mouse spoke truthfully. They were asked to provide a
brief justification for their answers. This allowed us to check whether the answers of the
participants are based on the right reasons and not on some irrelevant factors.5
To control for prosody, the test sentences were typed and shown to the
participants on the screen right after a story was completed. The participants read each
test sentence silently, and wrote down their answers on a score sheet. We reasoned that
when the participants are reading the sentences silently, they will freely assign a prosody
in order to make the sentence true. This reasoning relies on the assumption that listeners
follow the Gricean Maxim of Quality (Grice, 1975), and will give the speaker the benefit
of the doubt. Thus, if a participant rejects a statement as false, we conclude that her
grammar simply cannot generate the reading under investigation.
4 A reviewer asks about scope interaction between the subject QP and the object QP. In a separate
TVJT experiment, we tested whether 4 different quantifiers (daremo ‘every’, minna ‘every’, subete ‘every’,
futa-‘two’) in the object position can take scope over an existential quantifier (dareka ‘some’) in the subject
position. In this experiment, we were able to statistically confirm that these quantifiers in canonical
positions do exhibit scope rigidity. This study will be reported in the 18th International Congress of
Linguistics in Seoul, July 2008.
5 A reviewer notes that more follow-up questions should have been provided, for example asking the
participants to write the names of those that didn’t eat a watermelon in the scenario in Figure 1. In fact,
many participants gave such answers in their justification statements.
3.6 Findings
To obtain quantified results, whenever a participant accepted a statement, that response
was scored as one, and the rejection of a statement was scored as zero. The mean
percentage acceptance by condition was computed by first converting a participant’s
score out of four (= # of test trials) to a percentage, then arriving at the arithmetic mean
of the 32 (= # of participants per condition) percentage scores. So, if all 32 participants in
a condition accepted all 4 test trials, the percentage value for that condition is 100%. The
overall results from the experiment are summarized in Table 2 and in Figure 3.
Grammatical Function Scope Regular neg Wa-neg
Subject QP 2>neg 97.66% 93.75%
neg>2 9.38% 32.03%
Object QP 2>neg 85.94% 68.75%
neg>2 38.28% 64.84%
Table 2: Mean percentage acceptances by condition (Numeral quantifier study)
Each participant’s percentage score was entered into an analysis of variance
(ANOVA), which revealed the following effects. First, we found a main effect of scope
(F(7,248) = 157.182, p<.001): regardless of negation type and grammatical function,
participants were more likely to accept the 2>neg reading than the neg>2 reading.
Second, we found an interaction between grammatical function and scope (F(7,248) =
37.489, p<.001): independent of negation type, participants were significantly more likely
to accept the neg>2 reading on an object QP than they were on a subject QP. Third, we
found an interaction between negation type and scope (F(7,248) = 19.127, p<.001):
independent of grammatical function, participants were significantly more likely to
accept the neg>2 reading with -wa-negation than they were with regular negation.6
6 Participants were near perfect on filler items, indicating that they had no difficulty with the task, or
with negation or numeral futa in isolation.
Figure 3: Mean percentage acceptances by condition (Numeral quantifier study)
3.7 Discussion and analysis
In this section, we consider the two hypotheses in (9) against our findings and argue that
Hypothesis 1 is superior.
According to our findings, in ‘Subject QP -Regular neg’ conditions, while almost all
the speakers accepted the 2>neg reading (97.66%), speakers rarely accepted the neg>2
reading (9.38%). This fact supports our assumption that scope of negation and argument
QPs reflects structure: as the subject QP is high in the clause structure and it
asymmetrically c-commands negation, the 2>neg reading, but not the neg>2 reading, is
readily available to speakers. So, our findings regarding the scope of negation and a
subject QP are consistent with both Hypothesis 1 and Hypothesis 2, as in both
hypotheses, the subject QP is in [Spec,TP], asymmetrically c-commanding negation.
Further, the subject QP in [Spec,TP] does not reconstruct to [Spec,vP]. If it were to
reconstruct to [Spec,vP], then the object QP that has undergone a clause-internal
scrambling to the higher layer of vP should be able to take scope over it. But this is not
possible, as illustrated in (17). In (17), the object has moved out of VP, and the only
reading available is the one which the subject QP scopes over the object QP.
(17) Dareka on’nanoko-ga otokonoko daremo-o gakkoo-de hihanshi-ta.
some girl-NOM boy every-ACC school-in criticize-PST
‘Some girl criticized every boy in school.’ ( some>every, *every>some)
As scope rigidity holds between negation and the subject QP, we are justified in using
scope of negation and the object QP to determine the structural relation between the two,
to which we turn now.7
In ‘Object QP -2>neg -Regular neg’ condition, the majority of speakers accepted the
2>neg reading (85.94%). This suggests that a structure is available in which NegP is
hierarchically lower than the object QP, as in (9a). However, a structure in which NegP is
higher than the object QP, as in (9b), is also consistent with the data. This is because
numerals are typical indefinites that can take scope using the choice function strategy
(Reinhart, 1997). An indefinite as a choice function can take scope over any other
quantifiers in the same sentence. As the object QP in our test sentences contains a
numeral, futa (‘two’), speakers could have generated the 2>neg reading by Hypothesis 1,
or by Hypothesis 2 using the choice function strategy. So, the data in this condition
remains inconclusive in determining the superiority of the two hypotheses.
In ‘Object QP -neg>2 -Regular neg’ condition, the acceptance rate is only 38.28%.
This fact simply cannot be accounted for with Hypothesis 2, because this hypothesis can
7 A reviewer questions about the 9.38% acceptance rate in the ‘Subject QP -neg>2 -Regular neg’ con-
dition. Anticipating the discussion a bit, it might be that for those speakers that accepted the reading in this
condition, the verb may raise (along with negation) all the way up to C. In any case, we emphasize that the
fact that the majority of speakers rejected the reading in this condition supports our assumption that scope
rigidity holds between negation and argument QPs in Japanese. In languages that do not show scope
rigidity, such as English, a similar TVJT experiment in Musolino et al. (2000) showed that participants
almost always accepted both the Q>neg reading and the neg>Q reading in negative sentences containing a
subject universal quantifier.
only generate the neg>2 reading, hence incorrectly predicting near 100% acceptance
given this condition. This fact cannot be accounted for with Hypothesis 1 either, because
this hypothesis incorrectly predicts near 0% acceptance given this condition. The split
response, given this condition, however, can be accounted for with Hypothesis 1 in
conjunction with verb raising. According to Hypothesis 1, NegP is c-commanded by the
object QP, accounting for the half of the speakers who rejected the neg>2 reading. For
the other half of the speakers, a structure must be available in which negation is higher
than the object QP. This can be obtained if the verb raises to T0
, picking up negation on
its way to T0
, forming a complex head, as in (18b). Assuming the Kaynean definition of
c-command, as in (19), negation in 0
is able to c-command out of the complex head, and
take scope over the object.
(18) a. b.
(19) X c-commands Y iff X and Y are categories and X excludes Y [no segment of X
dominates Y] and every category that dominates X dominates Y (Kayne, 1994, p.
16).
But our reasoning so far implies that only half of the speakers have verb raising. If all the
speakers had verb raising, then all the speakers would incorrectly be predicted to accept
the neg>2 reading, just as the prediction made by Hypothesis 2.8
The idea that there may be a variation concerning verb-raising within a single speech
community has been proposed by Han et al. (2007). They report a similar TVJT
experiment on scope of negation and an object QP in Korean with similar split in scope
judgments. Han et al. argue that this split is a by-product of the head-final clausal
structure which obliterates string order evidence for verb raising. Thus, as far as verb
raising is concerned, they argue, the learners of Korean do not receive enough evidence
to decide the matter, and hence not all speakers acquire the same grammar. We take our
results from Japanese, another head-final language, to be a further support of the proposal
put forth in Han et al. and conclude that the neg>2 reading is available only to the
speakers that have acquired a verb raising grammar. This proposal is further supported by
the fact that the split in scope judgments is not within a speaker but between speakers:
that is, in general, a speaker that was tested for the neg>2 reading on an object QP and
regular negation either uniformly accepted or uniformly rejected the neg>2 reading,
giving rise to a bi-modal distribution of responses, as in Figure 4.9
8 A reviewer asks why the verb raising stops at T. In principle, verb raising all the way to C should be
possible. In fact, this may be what is going on for speakers who accept the neg>2 reading in subject
conditions, as entertained in footnote 7. But for us, the fact that the majority of participants did not accept
the neg>2 reading in subject conditions means that the verb stops at T for these speakers.
9 Interestingly, the literature on Japanese does not seem to agree on whether Japanese has verb raising, just
as there isn’t much agreement on scope judgments concerning negation and quantified arguments. For
instance, while Otani and Whitman (1991) and Koizumi (2000) argue for verb raising, Hoji (1998) points
out a flaw in Otani and Whitman’s argument essentially concluding that their argument is not sufficient to
support verb raising, and Fukui and Sakai (2003) refute all the arguments presented by Koizumi, and argue
for the position that Japanese has no verb raising. See Storoshenko (2004) for a summary of the literature
on this issue. A similar state of affairs for Korean is reported and summarized in Han et al. (2007).
Figure 4: Number of participants accepting neg>2 on an object QP with regular negation
(Numeral quantifier study)
We emphasize that the split in responses in ‘Object QP - neg>2 - Regular neg’ con-
dition cannot be attributed to contextual variation. In a TVJT experiment, the discourse
context of each trial is held constant for all participants given the same condition. Further,
when Lidz and Musolino (2002) conducted similar TVJT experiments with adult and
child speakers of English and Kannada to test the scope of negation and an object QP, the
pattern of responses was very consistent: in conditions testing the availability of neg>Q
reading in sentences containing negation and an object QP, the adult and the child partici-
pants in both languages almost always accepted the test sentences. The fact that
consistent responses were obtained from English and Kannada participants means that the
split responses of Japanese speakers on similar sentences are not a by-product of the task
itself, or some random variation in perception or comprehension. Rather, it is a reflection
of a real variation in the grammar of the speech community of Japanese.
We now need to recast the data in ‘Object QP - 2>neg - Regular neg’ condition in
light of the split in verb raising. For the speakers who do not have verb raising,
Hypothesis 1 straightforwardly generates the 2>neg reading. For the speakers who have
verb raising, the structure will make available the neg>2 reading, as negation
asymmetrically c-commands the object QP. But even these speakers will accept the
statements given because (i) the 2>neg reading can be generated by the choice function
strategy, as discussed earlier, and (ii) the vP-adjoined object c-commands negation,
according to the Kaynean definition of c-command. This then accounts for the higher
acceptance rate in this condition (85.94%) in comparison to the acceptance rate in ‘Object
QP - neg>2 - Regular neg’ condition (38.28%).
Let us now turn to -wa-negation conditions. Just as in ‘Subject QP - Regular neg’
conditions, in ‘Subject QP - Wa-neg’ conditions, participants more readily accepted the
2>neg reading (93.75%) than the neg>2 reading (32.03%). A one-way ANOVA,
comparing average scores across all eight conditions, and a post-hoc analysis using the
Tukey test did not yield a significant difference when the ‘Subject QP - 2>neg - Regular
neg’ condition was compared against the ‘Subject QP - 2>neg - Wa-neg’ condition, or
when ‘Subject QP - neg>2 - Regular neg’ condition was compared against ‘Subject QP -
neg>2 - Wa-neg’. This result is certainly consistent with the clause structure of
Hypothesis 1 where the subject, located high in the clause structure, is asymmetrically c-
commanding negation.
A one-way ANOVA and a post-hoc analysis using the Tukey test yielded no
significant difference between ‘Object QP - 2>neg - Regular neg’, ‘Object QP - 2>neg -
Wa-neg’ and ‘Object QP - neg>2 - Wa-neg’ conditions, when compared against each
other, whereas a significant difference was revealed when the ‘Object QP -neg>2 -
Regular neg’ was compared against each of the other three object conditions. That is,
even though the mean percentage acceptances in ‘Object QP - 2>neg - Wa-neg’ and
‘Object QP - neg>2 - Wa-neg’ conditions are not as high as the one in ‘Object QP -
2>neg - Regular neg’ condition, statistically, the acceptance rates of the two ‘Object QP -
Wa-neg’ conditions are not different from ‘Object QP - 2>neg - Regular neg’ condition.
However, the acceptance rates of the two ‘Object QP - Wa-neg’ conditions are
statistically different from ‘Object QP -neg>2 - Regular neg’ condition. Our analysis
therefore must freely allow both the 2>neg reading and the neg>2 reading in sentences
with an object QP and -wa-negation, just as we freely allowed the 2>neg reading in
sentences with an object QP and regular negation.10
To generate the neg>2 reading in ‘Object QP - neg>2 - Wa-neg’ condition, a structure
should be available for all speakers in which negation is higher than the object QP. How
could this be if negation is projected lower in the clause, as we just concluded from the
regular negation data? A clause structure containing wa-negation which is consistent with
Hypothesis 1 is as in (20), where -wa on the verb blocks the verb from coming together
with the tense inflection, and instead suru ‘do’ is inserted in T0
to support tense.
Exploiting the fact that -na has an affixal status morphologically, the neg>2 reading can
be generated if na moves and affixes onto suru in T0
. We assume that -na adjoins to T0
in
syntax and then it undergoes a morpho-phonological relinearization in morphology (in
the sense of Embick and Noyer 2001, and Fukui and Sakai 2003) obtaining the surface
string order suru+na+TENSE.
10
While it is not statistically significant, the relatively low acceptance rating for the two ‘Object - Wa-neg’
conditions, in comparison to ‘Object - 2>neg - Regular neg’ condition, remains in need of explanation. We
speculate that this is a result of the discourse-sensitivity of the -wa particle, and the fact that the test
sentences were presented without prosody. Even with the controlled discourse contexts of the TVJT trials, a
slight variation in prosody, as each participant was silently reading the test sentences with -wa, may have
led to a variation in responses. Presented with a fixed prosody for all participants, we expect that the
responses to these stimuli would be more stable. Moreover, although the difference between ‘Subject QP -
neg>2 - Regular neg’ and ‘Subject QP - neg>2 - Wa-neg’ conditions was not statistically significant, the
fact that the acceptance rating in the ‘Subject QP - neg>2 - Wa-neg’ condition is higher than in the ‘Subject
QP - neg>2 - Regular neg’ condition remains in need of explanation. A reviewer suggests that when topic
marker -wa is attached to a verb, it yields a contrastive topic meaning and when such a topic-marked verb
is negated, it is more likely to yield a negation wide scope reading. The same reviewer also suggests a
comparison between sentences with case particles on argument QPs and those with the topic marker -wa.
We certainly agree with the reviewer that the discourse sensitivity of -wa can lead to some interpretational
difference between the two types of sentences, and it would be interesting to do a TVJT experiment to tease
apart the difference. We leave this for future research.
(20)
The structure in (20) also generates the 2>neg reading in sentences with an object QP
and -wa-negation. This is because (i) as futa (‘two’) is a numeral, it can take scope over
negation using the choice function strategy, and (ii) the vP-adjoined object c-commands
negation, according to the Kaynean definition of c-command.
4 Comparison with the universal quantifier experiment
In this section, we discuss a similar study, presented in Han et al. (2004), which tested
the scopal relations between negation and the universal quantifier subete ‘every’ in object
position. We show that the results from this study further support our claims made above.
4.1 Methods and design
As in the previously-discussed experiment, we used a TVJT approach. For this
experiment we chose to concentrate solely on object QP’s, eliminating one variable from
the design. Thus, there were two factors with two levels each: negation (regular neg vs. -
wa neg) and scope ( >neg vs. neg> ), yielding four different conditions. Participants
were selected according to the same criteria as in the numeral quantifier experiment;
overall, 48 native Japanese speakers were recruited. We used a between-subjects design,
randomly assigning 12 participants to each group. The design is summarized in Table 3.11
Scope Regular neg Wa-neg
>neg n=12 n=12
neg> n=12 n=12
Table 3: Experimental design (Universal quantifier study)
11
There was no overlap between the set of participants in the universal quantifier experiment and the
numeral quantifier experiment.
4.2 Materials
For this experiment we used two types of test sentences, an object QP and regular
negation as in (21) and an object QP and -wa negation as in (22). Each test sentence was
used in a scenario compatible with the >neg reading, and a scenario compatible with the
neg>
reading.
(21) Donald-ga orenji subete-o tabe-na-katta.
Donald-NOM orange every-ACC eat-NEG-PST
‘Donald did not eat every orange.’
(22) Donald-ga orenji subete-o tabe-na-katta.
Donald-NOM orange every-ACC eat-NEG-PST
‘Donald did not eat every orange.’
For example, to test whether (21) and (22) have the >neg reading, we used a scenario
where Donald has three oranges, but does not eat any of them. To test for the neg>
reading, we used a scenario where Donald ate two out of the three oranges. Four test
trials were created for each condition. All test sentences for this experiment are listed in
Appendix C. To introduce variety into the responses, a series of unambiguously true and
unambiguously false filler trials were created, using both regular neg and -wa neg.
Similar to the numeral quantifier experiment, we also used quantifiers in subject
positions, and alternated the types of negation used between test and filler trials. A full
list of all filler sentences used in this experiment is presented in Appendix D.
4.3 Procedure
Participants were tested individually in this experiment, though the overall procedure is
the same as the experiment in section 3. All stimuli were presented as video recordings,
with participants giving true/false responses at the end of every trial. Sessions began with
four training trials, to familiarize participants with the task, followed by eight
experimental trials (four test and four filler) presented in a pseudo-random order.
In this experiment, all stimuli were presented aurally to the participants, rather than
typed on screen. The speaker providing the voice for Mickey Mouse was instructed to use
a neutral prosody for each sentence. However, any differences in responses between
participants cannot be ascribed to prosodic cues, as recordings were used; all participants
heard the same prosody.
4.4 Findings
Experimental findings are quantified in the same manner as the numeral quantifier experi-
ment. The percentage of acceptances for each condition is presented in Table 4, and
graphically in Figure 5.
Scope Regular neg Wa-neg
>neg 98% 98%
neg> 54% 94%
Table 4: Mean percentage acceptances by condition (Universal quantifier study)
Figure 5: Mean percentage acceptances by condition (Universal quantifier study)
The results here clearly show that the condition in which regular negation interacts
with an object universal quantifier is distinct from the other conditions, similar to the
pattern shown in the object QP conditions in the numeral quantifier experiment. To verify
this, a simple one-way ANOVA was carried out, comparing the scores across all four
conditions (F(3,44) = 9.156, p<0.0001). Further post-hoc pairwise testing confirms that
the neg> , regular negation results are significantly different from each of the other three
conditions. This is the exactly the same statistical pattern shown among the object QP
conditions in the numeral quantifier experiment.
4.5 Discussion and analysis
According to our findings, almost all speakers accepted the reading where the universal
quantifier scopes over regular negation, suggesting a low position for negation, in accor-
dance with Hypothesis 1. However, this result could also be compatible with the high
negation possibility presented in Hypothesis 2. This is due to the fact that a scenario
which satisfies the truth conditions for >neg will also satisfy the truth conditions for
neg> . That is, a situation in which Donald eats none of the oranges is also a situation in
which Donald eats ‘not all’ of the oranges. As such, the data from this condition are
inconclusive with respect to deciding between either of the two hypotheses. This is a well
known issue, identified as ‘the entailment issue’ by Lidz and Musolino (2002).
In ‘neg> -Regular neg’ condition, there was only slightly more than 50% acceptance.
This fact cannot be accounted for under Hypothesis 2, as the high negation structure will
only generate readings where negation scopes over the quantifier, predicting 100% accep-
tance. However, Hypothesis 1, with the low negation, would predict 0% acceptance.
Again, this result can be accounted for with an appeal to a split within the population
along the parameter of verb raising. For the half of the population for whom verb raising
is an option, wide scope negation is available, and the responses in this condition come
out to be true. For those with a non-verb-raising grammar, the quantifier takes the wider
scope, resulting in false responses to a neg> scenario.
This finding is further supported by the distribution of true and false responses within
the group, shown in Figure 6. Similar to the numeral quantifier study, an examination of
the responses within this group reveals a bi-modal distribution, with the majority of
speakers presenting an ‘all true’ or an ‘all false’ pattern. This indicates that the responses
here are being guided by a stable facet of the participants’ grammar, rather than varying
from scenario to scenario.
Figure 6: Number of participants accepting neg> on an object QP with regular negation
(Universal quantifier study)
Looking at the -wa negation cases, the data pattern here can be interpreted as compat-
ible with Hypothesis 1 with the structure in (20) in the same manner as in the discussion
around the numeral quantifier experiment. Almost all speakers accepted the neg>
reading and the >neg reading. The neg> reading is available because na is sitting high
in the clause structure, c-commanding the universally quantified object. The near 100%
acceptance rate in the ‘ >neg -Wa-neg’ condition is expected because a context that ver-
ifies the >neg reading also verifies the neg> reading and also the vP-adjoined object c-
commands negation, according to the Kaynean definition of c-command.
5 Conclusion
In this paper, we have shown, through scope judgments obtained from a numeral quanti-
fier TVJT experiment, that Japanese speakers uniformly accept the 2>neg reading, while
rejecting the neg>2 reading, in negative sentences with a subject QP. This fact supports
that the scope of negation and argument QPs reflects structure. We have also shown that
in negative sentences with an object QP, the majority of Japanese speakers seem to
accept the 2>neg reading with regular and -wa-negation, and neg>2 reading with -wa-
negation, but there is a split in the population when it comes to the neg>2 reading with
regular negation. We have attributed this split in scope judgments to a split in the
availability of verb raising within the population, and argued that all the data can be
accounted for if negation is projected lower in the clause structure. We have also shown
that these findings on the scope of negation and numeral QPs are consistent with the
findings from a previously published experiment (Han et al., 2004) on the scope of
negation and universal QPs. With our analysis, we are able to present a unified account of
the two types of sentential negation in Japanese, and also provide an explanation for the
disagreement in the extant literature on the topic of negation scope as a reflection of
speakers’ grammars. By supplementing traditional introspective judgments with a
quantitative analysis, we have been able to show that what appeared to be inconclusive
data consisting of conflicting judgments is reducible to a statistically significant pattern of
variation within the population, allowing us to gain new insight into the grammar, or
indeed grammars, of the language.
Appendix A: Test sentences - Numeral quantifier experi-
ment
This appendix provides the test sentences we used in each condition in the numeral
quantifier experiment.
(23) ‘Subject QP - 2>neg - Regular neg’; ‘Subject QP - neg>2 - Regular neg’
a. Futa-ri-no otokonoko-ga iwa-o kooen-de nage-na-katta.
two-CL-GEN boy-NOM rock-ACC park-in throw-NEG-PST
‘Two boys did not throw a rock in the park.’
b. Futa-ri-no on’nanoko-ga yubiwa-o omise-de kawa-na-katta.
two-CL-GEN girl-NOM ring-ACC shop-at buy-NEG-PST
‘Two girls did not buy a ring at the shop.’
c. Futa-ri-no otokonoko-ga suika-o shokudoo-de tabe-na-katta.
two-CL-GEN boy-NOM watermelon-ACC cafeteria-in eat-NEG-PST
‘Two boys did not eat a watermelon in the cafeteria.’
d. Futa-ri-no on’nanoko-ga inu-o omise-de nade-na-katta.
two-CL-GEN girl-NOM dog-ACC shop-in pet-NEG-PST
‘Two girls did not pet the dog in the shop.’
(24) ‘Subject QP - 2>neg - Wa-neg’; ‘Subject QP - neg>2 - Wa-neg’
a. Futa-ri-no otokonoko-ga iwa-o kooen-de nage-wa shi-na-katta.
two-CL-GEN boy-NOM rock-ACC park-in throw-TOP do-NEG-PST
‘Two boys did not throw a rock in the park.’
b. Futa-ri-no on’nanoko-ga yubiwa-o omise-de kai-wa shi-na-katta.
two-CL-GEN girl-NOM ring-ACC shop-at buy-wa do-NEG-PST
‘Two girls did not buy a ring at the shop.’
c. Futa-ri-no otokonoko-ga suika-o shokudoo-de tabe-wa shi-na-katta.
two-CL-GEN boy-NOM watermelon-ACC cafeteria-in eat-TOP do-NEG-PST
‘Two boys did not eat a watermelon in the cafeteria.’
d. Futa-ri-no on’nanoko-ga inu-o omise-de nade-wa shi-na-katta.
two-CL-GEN girl-NOM cat-ACC shop-in pet-TOP do-NEG-PST
‘Two men did not pet the cat in the pet shop.’
(25) ‘Object QP - 2>neg - Regular neg’; ‘Object QP - neg>2 - Regular neg’
a. Shigeru-san-ga futa-ri-no on’nanoko-o ie-no mae-de front-in dakishime-na-
Shigeru-san-NOM two-CL-GEN girl-ACC house-GEN hug-NEG-
katta.PST
‘Shigeru did not hug two girls in front of his house.’
b. Junko-ga futa-ri-no otokonoko-o rooka-de kera-na-katta.
Junko-NOM two-CL-GEN boy-ACC hallway-in kick-NEG-PST
‘Junko did not kick two boys in the hallway.’
c. John-ga futa-ri-no on’nanoko-o kooen-de mitsuke-na-katta.
John-NOM two-CL-GEN girl-ACC park-in find-NEG-PST
‘John did not find two girls in the park.’
d. Ken-ga futa-ri-no on’nanoko-o doobutsuen-de mochi-age-na-katta.
Ken-NOM two-CL-GEN girl-ACC zoo-in lift-up-NEG-PST
‘Ken did not lift up two girls in the zoo.’
(26) ‘Object QP - 2>neg - Wa-neg’; ‘Object QP - neg>2 - Wa-neg’
a. Shigeru-san-ga futa-ri-no on’nanoko-o ie-no mae-de dakishime-wa shi-na-
Shigeru-san-NOM two-CL-GEN girl-ACC house-GEN front-in hug-TOP do-NEG-
katta.PST
‘Shigeru did not hug two girls in front of his house.’
b. Junko-ga futa-ri-no otokonoko-o rooka-de keri-wa shi-na-katta.
Junko-NOM two-CL-GEN boy-ACC hallway-in kick-TOP do-NEG-PST
‘Junko did not kick two boys in the hallway.’
c. John-ga futa-ri-no on’nanoko-o kooen-de mitsuke-wa shi-na-katta.
John-NOM two-CL-GEN girl-ACC park-in find-TOP do-NEG-PST
‘John did not find two girls in the park.’
d. Ken-ga futa-ri-no on’nanoko-o doubutsuen-de mochi-age-wa shi-na-katta.
Ken-NOM two-CL-GEN girl-ACC zoo-in lift-up-TOP do-NEG-PST
‘Ken did not lift up two girls in the zoo.’
Appendix B: Filler sentences - Numeral quantifier experi-
ment
This appendix provides the filler sentences we used in each condition in the numeral
quantifier experiment.
(27) ‘Subject QP - 2>neg - Regular neg’; ‘Subject QP - neg>2 - Regular neg’
a. Kaba-ga kirikabu-ni nobori-wa shi-na-katta.
hippo-NOM tree trunk-onto climb-TOP do-NEG-PST
‘The hippo did not climb onto the tree trunk.’
b. Dora-ga kagami-o omise-de kai-wa shi-na-katta.
Dora-NOM mirror-ACC shop-at buy-TOP do-NEG-PST
‘Dora did not buy the mirror at the shop.’
c. Lisa-ga futa-ri-no otokonoko-o ie-no mae-de dakishime-ta.
Lisa-NOM two-CL-GEN boy-ACC house-GEN front-in hug-PST
‘Lisa hugged two boys in front of her house.’
d. On’nanohito-ga futa-ri-no otokonoko-o paati-ni manei-ta.
woman-NOM two-CL-GEN boy-ACC party-to invite-PST
‘The woman invited two boys.’
(28) ‘Subject QP - 2>neg - Wa-neg’; ‘Subject QP - neg>2 - Wa-neg’
a. Kaba-ga kirikabu-ni nobora-na-katta.
hippo-NOM tree trunk-onto climb-NEG-PST.
‘The hippo did not climb onto the tree trunk.’
b. Dora-ga kagami-o omise-de kawa-na-katta.
Dora-NOM mirror-ACC store-at buy-NEG-PST
‘Dora did not buy the mirror at the store.’
c. Lisa-ga futa-ri-no otokonoko-o ie-no mae-de dakishime-ta.
Lisa-NOM two-CL-GEN boy-ACC house-GEN front-in hug-PST
‘Lisa hugged two boys in front of her house.’
d. On’nanohito-ga futa-ri-no otokonoko-o paati-ni manei-ta.
woman-NOM two-CL-GEN boy-ACC party-to invite-PST
‘The woman invited two boys.’
(29) ‘Object QP - 2>neg - Regular neg’; ‘Object QP - neg>2 - Regular neg’
a. Kaba-ga kirikabu-ni nobori-wa shi-na-katta.
hippo-NOM tree trunk-onto climb-TOP do-NEG-PST
‘The hippo did not climb onto the tree trunk.’
b. Dora-ga kagami-o omise-de kai-wa shi-na-katta.
Dora-NOM mirror-ACC shop-at buy-TOP do-NEG-PST
‘Dora did not buy the mirror at the shop.’
c. Futa-ri-no otoko-ga iwa-o kooen-de nage-ta.
Two-CL-GEN man-NOM rock-ACC park-in throw-PST
‘Two men threw a rock.’
d. Futa-ri-no otoko-ga neko-oomise-de nade-ta.
two-CL-GEN man-NOM cat-ACC shop-in pet-PST
‘Two men petted the cat in the shop.’
(30) ‘Object QP - 2>neg - Wa-neg’; ‘Object QP - neg>2 - Wa-neg’
a. Kaba-ga kirikabu-ni nobora-na-katta.
hippo-NOM tree trunk-onto climb-NEG-PST.
‘The hippo did not climb onto the tree trunk.’
b. Dora-ga kagami-o omise-de kawa-na-katta.
Dora-NOM mirror-ACC store-at buy-NEG-PST
‘Dora did not buy the mirror at the store.’
c. Futa-ri-no otoko-ga iwa-o kooen-de nage-ta.
Two-CL-GEN man-NOM rock-ACC park-in throw-PST
‘Two men threw a rock in the park.’
d. Futa-ri-no otoko-ga neko-o omise-de nade-ta.
two-CL-GEN man-NOM cat-ACC shop-in pet-PST
‘Two men petted the cat in the shop.’
Appendix C: Test sentences - Universal quantifier experi-
ment
This appendix provides the test sentences we used in each condition in the universal
quantifier experiment.
(31) ‘ >neg - Regular neg’; ‘neg> - Regular neg’
a. Lisa-ga yubiwa-subete-o kaw-ana-katta.
Lisa-NOM ring-every-ACC buy-NEG-PST
‘Lisa did not buy every ring.’
b. Donarudo-ga orenji-subete-o tabe-na-katta.
Donald-NOM orange-every-ACC eat-NEG-PST
‘Donald did not eat every orange.’
c. Otokonohito-ga kuruma-subete-o araw-ana-katta.
man-NOM car-every-ACC wash-NEG-PST
‘The man did not wash every car.’
d. Otokonohito-ga zou-subete-o mochi-age-na-katta.
man-NOM elephant-every-ACC lift-up-NEG-PST
‘The man did not lift up every elephant.’
(32) ‘ >neg - Wa-neg’; ‘neg> - Wa-neg’
a. Lisa-ga yubiwa-subete-o kai-wa shi-na-katta.
Lisa-NOM ring-every-ACC buy-TOP do-NEG-PST
‘Lisa did not buy every ring.’
b. Donarudo-ga orenji-subete-o tabe-wa shi-na-katta.
Donald-NOM orange-every-ACC eat-TOP do-NEG-PST
‘Donald did not eat every orange.’
c. Otokonohito-ga kuruma-subete-o arai-wa shi-na-katta.
man-NOM car-every-ACC wash-TOP do-NEG-PST
‘The man did not wash every car.’
d. Otokonohito-ga zou-subete-o mochi-age-wa shi-na-katta.
man-NOM elephant-every-ACC lift-up-TOP do-NEG-PST
‘The man did not lift up every elephant.’
Appendix D: Filler sentences - Universal quantifier
experiment
This appendix provides the filler sentences we used in each condition in the universal
quantifier experiment.
(33) ‘ >neg - Regular neg’; ‘neg> - Regular neg’
a. Zou-ga kirikabu-ni nobori-wa shi-na-katta.
elephant-NOM tree trunk-onto climb-TOP do-NEG-PST
‘The elephant did not climb onto the tree trunk.’
b. On’nanohito-ga Lisa-ni kagami-o uri-wa shi-na-katta.
woman-NOM Lisa-to mirror-ACC sell-TOP do-NEG-PST
‘The woman did not sell the ring to Lisa.’
c. Otokonohito-subete-ga iwa-o nage-ta.
man-every-NOM rock-ACC throw-PST
‘Every man threw a rock.’
d. Kyooryuu-subete-ga kirikabu-kara ochi-ta.
dinosaur-every-NOM tree trunk-from fall-PST
‘Every dinosaur fell from the tree trunk.’
(34) ‘ >neg - Wa-neg’; ‘neg> - Wa-neg’
a. Zou-ga kirikabu-ni nobor-ana-katta.
elephant-NOM tree trunk-onto climb-NEG-PST
‘The elephant did not climb onto the tree trunk.’
b. On’nanohito-ga Lisa-ni kagami-o ur-ana-katta.
woman-NOM Lisa-to mirror-ACC sell-NEG-PST
‘The woman did not sell the ring to Lisa.’
c. Otokonohito-subete-ga iwa-o nage-ta.
man-every-NOM rock-ACC throw-PST
‘Every man threw a rock.’
d. Kyooryuu-subete-ga kirikabu-kara ochi-ta.
dinosaur-every-NOM tree trunk-from fall-PST
‘Every dinosaur fell from the tree trunk.’
References
Crain, Stephen, and Cecile McKee. (1985). The acquisition of structural restrictions on
anaphora. In Proceedings of NELS 16. GLSA.
Crain, Stephen, and Rosalind Thornton. (1998). Investigations in universal grammar: A
guide to research in acquisition of syntax and semantics. Cambridge, Massachusetts:
The MIT Press.
Embick, David, and Rolf Noyer. (2001). Movement operations after syntax. Linguistic
Inquiry 32:555–595.
Fukui, Naoki, and Hiromu Sakai. (2003). The visibility guideline for functional
categories: Verb raising in Japanese and related issues. Lingua 113:321–375.
Grice, Herbert Paul. (1975). Logic and conversation. In Speech acts, ed. P. Cole and J. L.
Morgan, volume III of Syntax and Semantics, 41–58. NY: Academic Press.
Han, Chung-hye, Jeffrey Lidz, and Julien Musolino. (2007). Verb-raising and grammar
competition in Korean: Evidence from negation and quantifier scope. Linguistic
Inquiry 38:1–47.
Han, Chung-hye, Dennis Ryan Storoshenko, and Yasuko Sakurai. (2004). Scope of
negation, and clause structure in Japanese. In Proceedings of the 30th Annual Meeting
of the Berkeley Linguistics Society. Berkeley: Berkeley Linguistics Society.
Hoji, Hajime. (1985). Logical form constraints and configurational structure in Japanese.
Doctoral Dissertation, University of Washington.
Hoji, Hajime. (1998). Null object and sloppy identity in Japanese. Linguistic Inquiry
29:127– 152.
Kayne, Richard. (1994). The antisymmetry of syntax, volume 25 of Linguistic Inquiry
Monograph. Cambridge, London: The MIT Press.
Koizumi, Masatoshi. (2000). String vacuous overt verb raising. Journal of East Asian
Linguistics 9:227–285.
Kuno, Susumu. (1973). The structure of the Japanese language. Cambridge, MA: MIT
Press.
Kuno, Susumu. (1980). The scope of question and negation in some verb-final languages.
In Proceedings of the Chicago Linguistic Society, ed. Jody Kreiman and Almerindo
E. Ojeda, volume 16, 155–169. Chicago Linguistic Society.
Kuroda, Shige-Yuki. (1979). Generative grammatical studies in Japanese language. New
York, London: Garland Publishing.
Lidz, Jeffrey, and Julien Musolino. (2002). Children’s command of quantification.
Cognition 84:113–154.
Miyagawa, Shigeru. (2001). The EPP, scrambling and wh-in-situ. In Ken hale: A life in
language, 293–338. Cambridge, MA: MIT Press.
Miyagawa, Shigeru. (2002). The nature of weak islands. Conference handout, KGGC
Summer School.
Musolino, Julien, Stephen Crain, and Rosalind Thornton. (2000). Navigating negative
quantificational space. Linguistics 38:1–32.
Nakanishi, Kimiko. (2006). Even, only, and negative polarity in Japanese. In Proceedings
of Semantics and Linguistic Theory (SALT) 16.
Oishi, Hitomi. (1986). Scope interpretation of negation in Japanese. Doctoral
Dissertation, University of Michigan.
Ota, Akira, and Yasuhiko Kato. (1986). The relative scope of neg and quantifiers in
English and Japanese. Sophia Linguistica 20-21:25–40.
Otani, Kazuyo, and John Whitman. (1991). V-raising and VP-ellipsis. Linguistic Inquiry
22:345–358.
Papafragou, Anna, and Julien Musolino. (2003). Scalar implicatures: Experiments at the
semantics-pragmatics interface. Cognition 86:253–282.
Reinhart, Tanya. (1997). Quantifier scope: how labor is divided between QR and choice
function. Linguistics and Philosophy 20:335–397.
Rizzi, Luigi. (1990). Relativized minimality. Cambridge, MA: MIT Press.
Storoshenko, Dennis Ryan. (2004). Negation scope and phrase structure in Japanese.
Master’s thesis, Simon Fraser University.
Yatabe, Shuichi. (1996). Negation and focusing in the grammar of Japanese. In Studies
on the universality of constraint-based phrase structure grammars, ed. Takao Gunji,
217–225. Osaka, Japan: Osaka University.
Chung-hye Han, Dennis Ryan Stroshenko and Yasuko Sakurai
Department of Linguistics, Simon Fraser University
8888 University Drive, Burnaby BC
Canada, V5A 1S6
{chunghye,dstorosh,ysakurai}@sfu.ca